Unsupervised deep learning of landscape typologies from remote sensing images and other continuous spatial data. (September 2022)

• Record Type: Journal Article
• Authors: van Strien, Maarten J.; Grêt-Regamey, Adrienne
• Is Part Of: Environmental modelling & software. Volume 155(2022)
• Publication Date: 2022-09
• Subjects: Unsupervised clustering -- Image clustering -- Landscape planning -- Convolutional neural networks; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2022.105462
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: The identification of landscape classes facilitates the implementation of planning strategies. Although landscape patterns are key distinctive features of landscape classes, existing unsupervised clustering techniques for clustering landscapes rely on categorical input data to quantify such patterns and consider only a limited number of pattern metrics. To unlock the great potential of continuous spatial data, such as remote sensing images, for generating landscape typologies, we adapted a novel unsupervised deep learning method (Deep Convolutional Embedded Clustering; DCEC) to generate a landscape typology for Switzerland. DCEC encodes lower-dimensional representations of input images in a hidden layer, which is simultaneously used to divide the images into well-distinguishable clusters. We applied DCEC to image tiles extracted from satellite images as well as ecological, demographic and terrain layers. DCEC successfully distinguished 45 landscape classes in the continuous input data. We conclude that DCEC is a promising new method in landscape and land-system research. Highlights: Landscape patterns are a key distinguishing property of landscape types. Our unsupervised deep learning approach clusters images based on learned patterns. Our approach can be applied directly to continuous data, such as satellite images. The range of considered patterns surpasses that of traditional landscape metrics. The cluster quality is better than that of traditional landscape clustering methods.